Hyperammonemia in gene-targeted mice lackingfunctional hepatic glutamine synthetaseNatalia Qvartskhavaa,1, Philipp A. Langa,b,1, Boris Görga, Vitaly I. Pozdeeva, Marina Pascual Ortiza, Karl S. Langc,Hans J. Bidmond, Elisabeth Langa, Christina B. Leibrocke, Diran Herebianf, Johannes G. Bodea, Florian Lange,and Dieter Häussingera,2

aDepartment of Gastroenterology, Hepatology, and Infectious Diseases, bDepartment of Molecular Medicine II, dCécile and Oskar Vogt Institute for BrainResearch, and fDepartment of General Pediatrics, Neonatology, and Pediatric Cardiology, Heinrich Heine University Düsseldorf, 40225 Düsseldorf, Germany;cInstitute of Immunology, Medical Faculty, University of Duisburg-Essen, 45147 Essen, Germany; and eDepartment of Physiology, University of Tübingen,72076 Tübingen, Germany

Edited by Arthur J. L. Cooper, New York Medical College, Valhalla, NY, and accepted by the Editorial Board March 16, 2015 (received for review December15, 2014)

Urea cycle defects and acute or chronic liver failure are linked tosystemic hyperammonemia and often result in cerebral dysfunctionand encephalopathy. Although an important role of the liver inammonia metabolism is widely accepted, the role of ammoniametabolizing pathways in the liver for maintenance of whole-bodyammonia homeostasis in vivo remains ill-defined. Here, we show bygeneration of liver-specific Gln synthetase (GS)-deficient mice thatGS in the liver is critically involved in systemic ammonia homeostasisin vivo. Hepatic deletion of GS triggered systemic hyperammonemia,which was associated with cerebral oxidative stress as indicated byincreased levels of oxidized RNA and enhanced protein Tyr nitra-tion. Liver-specific GS-deficient mice showed increased locomo-tion, impaired fear memory, and a slightly reduced life span. Inconclusion, the present observations highlight the importance ofhepatic GS for maintenance of ammonia homeostasis and establishthe liver-specific GS KO mouse as a model with which to studyeffects of chronic hyperammonemia.

hepatic encephalopathy | metabolic zonation | oxidative stress |RNA oxidation | glutamine

Hepatic ammonia and Gln metabolism are embedded into astructural/functional organization in the liver acinus, which

allows efficient ammonia detoxification by the liver (1, 2). Hepaticurea synthesis in periportal hepatocytes is dependent on the ac-tivity of carbamoylphosphate synthetase, the rate-controlling en-zyme of the urea cycle requiring ammonia as a substrate (3). Theaffinity of carbamoylphosphate synthetase for ammonia is low, andadequate flux through the urea cycle requires the establishment ofhigh ammonia concentrations in periportal hepatocytes, which isachieved through ammonia amplification by periportal glutamin-ase activity (1–3). Excess ammonia not used by urea synthesis istaken up with high affinity by a small perivenous hepatocytepopulation (so-called “perivenous scavenger cells”) (1), which de-toxifies ammonia by amidation of Glu (4). This reaction is cata-lyzed by Gln synthetase (GS), which is specifically expressed in thissmall population of hepatocytes surrounding the terminal hepaticvenule but not in other hepatocytes (4–6). This high-affinity am-monia removal by perivenous hepatocytes thus prevents spilloverof hepatic ammonia into the systemic circulation. The acinarcompartmentation of urea synthesis, Gln hydrolysis, and Gln syn-thesis therefore allows the generation of sizable ammonia con-centrations in hepatic tissue needed for urea formation without therisk of toxic ammonia concentrations in systemic circulation (1).Ammonia is toxic, particularly to the brain, where it can trigger

hepatic encephalopathy (HE). HE is seen as the clinical mani-festation of a low-grade cerebral edema with oxidative/nitrosativestress and subsequent derangements of signal transduction, neu-rotransmission, synaptic plasticity, and oscillatory networks in thebrain (7–9). In particular, the oxidative/nitrosative stress responseresults in protein Tyr nitration (PTN) and RNA oxidation (9, 10).

Astrocytes in close proximity to the blood–brain barrier exhibitstrong PTN, possibly affecting its permeability (11). Data derivedfrom HE animal models suggest a relationship between impairedfunctions of brain regions involved in cognition, learning, memoryformation, and motor control, as well as elevated markers foroxidative stress in the hippocampus (12, 13), cerebral cortex (14),and cerebellum (15, 16). Consistently, increased markers for oxi-dative stress, such as PTN and RNA oxidation, have also beenshown in postmortem human brain tissue of patients with livercirrhosis and HE (17). In mice, ammonia can also inhibit potas-sium buffering by astrocytes, which results in increased extracel-lular potassium and may contribute to neuronal depolarizationand dysfunction, and, consequently, altered behavior (18).Liver damage can trigger defects in the ammonia metabolizing

pathways, which consequently increase ammonia levels in circu-lating blood (19, 20). For example, LPS-induced liver injury resultsin PTN of hepatic GS, which inactivates the enzyme (21). Theimpact of the urea cycle for ammonia metabolism can be seen inchildren with urea cycle defects, who exhibit hyperammonemiaand cognitive symptoms (22). In turn, hereditary GS deficiency inhumans is a rare disorder leading to Gln deficiency and severedisease with a lethal outcome (23). Although destruction of peri-venous hepatocytes by carbon tetrachloride intoxication impairsammonia detoxification in perfused rat liver (24), the in vivo sig-nificance of the GS in perivenous hepatocytes has remained elu-sive. Because Gln can be synthesized in a wide variety of tissues

Significance

Ammonia metabolism in the liver is critical to prevent seriousclinical conditions, such as hepatic encephalopathy. It was hy-pothesized that the Gln synthetase (GS) canmetabolize ammoniawith high affinity in the perivenous region of the liver. However,the in vivo relevance of this metabolic pathway remains unclearin view of other intra- and extrahepatic ammonia metabolizingpathways. Here, we show by creating a conditional GS KOmouse that specific deletion of the GS in the liver results inincreased ammonia levels in the blood, induction of oxidativestress in brain tissue, and behavior abnormalities. In conclu-sion, GS in the liver is a key player in the maintenance ofammonia homeostasis.

Author contributions: D. Häussinger designed research; N.Q., P.A.L., B.G., V.I.P., M.P.O.,K.S.L., H.J.B., E.L., C.B.L., and D. Herebian performed research; J.G.B., F.L., and D. Häussingeranalyzed data; and N.Q., P.A.L., B.G., and D. Häussinger wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission. A.J.L.C. is a guest editor invited by theEditorial Board.1N.Q. and P.A.L. contributed equally to this work.2To whom correspondence should be addressed. Email: haeussin@uni-duesseldorf.de.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1423968112/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1423968112 PNAS | April 28, 2015 | vol. 112 | no. 17 | 5521–5526

PHYS

IOLO

GY

Dow

nloa

ded

by g

uest

on

July

15,

202

0

(4), specific deletion of hepatic GS would not be expected to resultin Gln deficiency. However, lack of hepatic GS may disruptperivenous ammonia metabolism, thus leading to increasedammonia concentrations in systemic blood.To test this hypothesis, gene-targeted mice lacking functional

hepatic GS were generated and analyzed. As a result of liver-specific GS deletion, mice developed systemic hyperammonemia,which was accompanied by cerebral RNA oxidation, PTN, motoricand behavioral abnormalities, and a reduced life span.

ResultsEffect of Liver-Specific GS Deletion on Liver Tissue Integrity, Zonation,and Systemic Ammonia Levels. Liver injury can result in a reductionof hepatic GS activity (21, 24). Because other potent ammoniadetoxicating mechanisms exist in the body, the role of the GS inperivenous liver cells remains controversial. To investigate the roleof GS, we created conditional GS KO mice (Fig. 1A), which werecrossed with a mouse strain where Cre is expressed under thealbumin promotor (25). As expected, GS was strongly expressed inliver tissue in locus of X-over P1 (loxP) sites flanked by glutamate-

ammonia ligase gene (Glulfl/fl) mice in the absence of Crerecombinase but was undetectable in Glulfl/fl × albumin promoter-controlled Cre recombinase gene (Alb-Cre+) mice (Fig. 1 B andC). Deletion of GS in the liver did not affect GS expression in thebrain or in muscle tissue (Fig. 1 D and E). GS activity was high inliver tissue from Glulfl/fl mice, whereas only residual GS activitywas detected in liver tissue from Glulfl/fl × Alb-Cre+ mice (Fig. 1F),which is probably attributable to tissue resident macrophages (26).Next, we investigated whether hepatic deletion of GS affectedliver architecture or damage. Histologically, no difference betweenWT and Glulfl/fl × Alb-Cre+ mice in snap-frozen liver sections wasobserved (Fig. 2A). Furthermore, activities of Asp and Ala ami-notransferases in serum were not different between WT and KOmice, indicating that the absence of hepatic GS did not result inliver damage (Fig. 2B). Although GS was selectively deleted inperivenous hepatocytes (Fig. 2C), ornithine aminotransferase(OAT), as well as Rhesus family B glycoprotein, which showsimilar acinar localization as GS, were still detectable in thesespecialized cells (27, 28) (Fig. 2 C and D). These data indicatedthat the zonal organization of the liver was not affected by theliver-specific GS KO. However, when we checked blood ammonialevels, strongly elevated ammonia concentrations were observed inblood plasma from KO mice compared with WT mice (Fig. 2E).This increase in plasma ammonia concentration was probably notdue to a urea cycle defect, because expression of genes encodingfor enzymes of the urea cycle in the liver was similar in WT andGlulfl/fl × Alb-Cre+ mice (Fig. S1A). Moreover, serum urea nitro-gen concentrations did not differ between WT and Glulfl/fl × Alb-Cre+ mice (Fig. S1B). Furthermore, serum Gln, Glu, and Ala

B

D

GAPDH

+ + –

+

loxP-Glul

Alb-Cre+–

liver

GS

+ + –

++–

brain

GAPDH

loxP-Glul

Alb-Cre

GS

F

GS

activ

ity (a

rbitr

ary

units

) Glulfl/fl

Glulfl//fl x Alb-Cre+

p < 0.01

0.00

0.02

0.04

0.06

A

C

GS

GSHoechst

Glulfl/flGlulfl/fl x

Alb-Cre+

GS

GSF-actin

Glulfl/flGlulfl/fl x

Alb-Cre+E

GS

GAPDH

+ +

+–

loxP-Glul

Alb-Cre

muscle

Fig. 1. Organ-specific deletion of GS in the Glulfl/fl × Alb-Cre+ mice. (A) Gene-targeted mice lacking functional hepatic GS were generated as describedin Materials and Methods. (B) GS and GAPDH expression in liver tissue fromGlulfl/fl × Alb-Cre+, Alb-Cre+, and Glulfl/fl mice is shown using Western blotanalysis. One representative of n = 3 is shown. (C) Snap-frozen sections fromliver tissue harvested from Alb-Cre+, Glulfl/fl, and Glulfl/fl × Alb-Cre+ mice werestained with anti-GS (green) and F-actin (red) antibodies. One representativeof n = 3 is shown. (Scale bar: 200 μm.) (D, Left) Western blot analysis for GS andGAPDH on protein samples obtained from brain tissue from Glulfl/fl × Alb-Cre+,Alb-Cre+, and Glulfl/fl mice was performed. (D, Right) GS expression andGAPDH expression were assessed in muscle tissue of Glulfl/fl × Alb-Cre+ andGlulfl/fl mice by Western blot analysis. One representative of n = 3 is shown.(E) Immunostaining of snap-frozen sections from brain tissue harvested fromAlb-Cre+, Glulfl/fl, and Glulfl/fl × Alb-Cre+ mice with an anti-GS antibody (green)and Hoechst 34580 (blue) was performed. One representative of n = 3 isshown. (Scale bar: 50 μm.) (F) GS activity was assessed in liver tissue harvestedfrom Glulfl/fl and Glulfl/fl × Alb-Cre+ mice (n = 3–7, respectively).

Enzy

me

activ

ity (U

/L)

AST ALT

Am

mon

ia (µ

mol

/L)

8-9 weeks

GlulGlul x Alb-Cre+

p < 0.001

n.s. n.s.

Glul

Glul x Alb-Cre+

H&E

GS

OAT-1

Hoechst

MergeRhBG

HoechstRhBG

GlulGlul x Alb-Cre+

GlulGlul x Alb-Cre+Glul

Glul x Alb-Cre+

0

50

100

150

200

0

125

250 p < 0.05

12-14 months

A B

E

C

D

Fig. 2. Intact liver architecture/zonation and elevated systemic ammonialevels by liver-specific deletion of GS. (A) Representative H&E-stained sectionsof snap-frozen liver tissue obtained from Glulfl/fl × Alb-Cre+ and Glulfl/fl mice ofn = 3 is shown. (Scale bar: 100 μm.) (B) Asp aminotransferase (AST) and Alaaminotransferase (ALT) activity was determined in the serum ofGlulfl/fl (n = 10)and Glulfl/fl × Alb-Cre+ mice (n = 12). n.s., not statistically significantly different.(C ) Immunofluorescence analyses of snap-frozen liver tissue from Glulfl/fl ×Alb-Cre+ (Lower) andGlulfl/fl (Upper) mice were performed for GS (red), ornithineaminotransferase (OAT; green), and Hoechst 34580 (blue). One representativeset of images of n = 3 is shown. (Scale bar: 20 μm.) (D) Immunofluorescenceanalyses of snap-frozen liver tissue from Glulfl/fl × Alb-Cre+ (Right) and Glulfl/fl

(Left) mice were performed for ammonia transporter Rh family B glycoprotein(RhBG; red) and Hoechst 34580 (gray). One representative set of images of n = 3(Glulfl/fl) and n = 4 (Glulfl/fl ×Alb-Cre+) is shown. (Scale bar: 200 μm.) (E) Ammonialevels were assessed in blood samples collected by cardiac puncture from 8- to9-wk-old Glulfl/fl ×Alb-Cre+ mice and Glulfl/fl mice (Left, n = 7–8, respectively) andfrom 12- to 14-month-old animals (Right, n = 3).

5522 | www.pnas.org/cgi/doi/10.1073/pnas.1423968112 Qvartskhava et al.

Dow

nloa

ded

by g

uest

on

July

15,

202

0

concentrations were not significantly different betweenGlulfl/fl andGlulfl/fl × Alb-Cre+ mice (Fig. S1C). Taken together, these dataindicate that liver-specific deletion of GS does not affect liverarchitecture but allows ammonia spillover into the systemic cir-culation, thereby leading to hyperammonemia.

Liver-Specific GS KO Triggers PTN and RNA Oxidation in Mouse Brain.Ammonia intoxication and HE were shown to trigger PTN nitrationand RNA oxidation in the rat and human brain, respectively (10,17, 21, 29). Therefore, we analyzed whether these effects could alsobe observed in brain tissue from liver-specific GS KO mice.Whereas PTN was enhanced in the cerebellum, hippocampus, andsomatosensory cortex of Glulfl/fl × Alb-Cre+ mice (Fig. 3 A–C), itwas similar to PTN in WT mice in the piriform cortex (Fig. 3D).Thus, nitration of proteins due to chronic ammonia intoxication

affects specific regions of the brain (Fig. 3E). Next, we investigatedwhether RNA oxidation was present in brain tissue using im-munofluorescence. As evidenced by anti–8-hydroxy-2-(de)oxy-guanosine [8-OH(d)G] immunoreactivity, RNA oxidation wasconsistently found in the cerebellum, hippocampus, and somato-sensory cortex (Fig. 4 A–C) but not in the piriform cortex (Fig.4D). When brain slices were treated with RNase, no anti–8-OH(d)Gimmunoreactivity was detectable on the brain slices, indicatingthat the oxidation was specific for RNA (Fig. S2). These resultswere also confirmed with Northwestern blotting (Fig. 4E and Fig.S3 A–D). Interestingly, in the cerebellum of mice lacking hepaticGS, we identified exceptionally high levels of oxidized RNA inPurkinje cells (Fig. S4), which critically control motor functions(30). Taken together, specific deletion of GS in the liver results inincreased ammonia levels and, consequently, biochemical changesin brain tissue of Glulfl/fl × Alb-Cre+ mice.

Glulfl/fl × Alb-Cre+ Mice Show No Microglia Activation in MouseCerebral Cortex. Microglia activation was shown in the brain ofanimal models for acute and chronic liver failure (16, 31) and in

human postmortem brain samples from patients with liver cir-rhosis and HE (32). However, microglia activation markers, suchas ionized calcium-binding adaptor protein 1 (Iba1) expression,and microglia morphology, were unchanged in the cerebral cortexof Glulfl/fl × Alb-Cre+ mice compared with WT mice (Fig. 5A). Inline with this finding, cerebrocortical Iba1 and CD14 mRNA (Fig.5B) and protein (Fig. S5A) levels were unchanged inGlulfl/fl × Alb-Cre+ mice compared with Glulfl/fl mice. In the brain, proin-flammatory cytokines are largely synthesized by activated micro-glia (33). Also, mRNA (Fig. 5C) and protein (Fig. S5B) levels ofthe proinflammatory cytokines IL-1β, IL-6, and TNF-α remainedunchanged in the cerebral cortex of Glulfl/fl × Alb-Cre+ micecompared with Glulfl/fl mice. These data suggest that systemichyperammonemia inGlulfl/fl ×Alb-Cre+ mice is not associated withmicroglia activation and increased synthesis of proinflammatorycytokines in cerebral cortex.

Glulfl/fl × Alb-Cre+ Mice Exhibit Behavioral Abnormalities. Changes inammonia levels in patients can trigger signs of HE. We thereforetested whether liver-specific deletion of GS resulted in behav-ioral abnormalities. Interestingly, when we monitored animalactivity using light barrier-equipped cages, increased activity of

0.0

0.8

1.6

C H S.C.Rela

tive

Prot

ein-

Tyr n

itrat

ion

(fo

ld in

duct

ion

/WT)

P.C.

p < 0.001

p < 0.01p < 0.01

n.s.

A DCCerebellum

SomatosensoryCortex

PiriformCortex

BHippocampus

EGlulfl/fl

Glulfl//fl x Alb-Cre+

NO2-Tyr

GAPDH

150

10075

5037

25

15

10

150

10075

5037

25

15

10

250150100

5037

25

15

10

75

250150100

5037

25

15

10

75

Glulfl/fl Glulfl//fl

x Alb-Cre+Glulfl/fl Glulfl//fl

x Alb-Cre+Glulfl/fl Glulfl//fl

x Alb-Cre+Glulfl/fl Glulfl//fl

x Alb-Cre+

Fig. 3. Hepatic GS KO triggers PTN in mouse brain. Protein samples harvestedfrom brain slices of the cerebellum, hippocampus, somatosensory cortex, andpiriform cortex ofGlulfl/fl and Glulfl/fl ×Alb-Cre+mice were tested for PTN usingWestern blot analysis. (A–D, Upper) Representative blots showing anti–3′-nitrotyrosine immunoreactivity. (A–D, Lower) GAPDH served as a loadingcontrol. (E) Mean ± SEM of the densitometry of the cerebellum (C), hippo-campus (H), somatosensory cortex (S.C.), and cortex piriform (C.P.) is presented(n = 6 for Glulfl/fl and n = 9 for Glulfl/fl × Alb-Cre+ mice for the cerebellum andhippocampus, n = 6 for Glulfl/fl and n = 8 for Glulfl/fl × Alb-Cre+ mice for thesomatosensory cortex, and n = 3 for Glulfl/fl and n = 6 for Glulfl/fl × Alb-Cre+

mice for the cortex piriform).

B

E

Cerebellum Hippocampus

Piriform CortexSomatosensory cortex

0

1

2

3

Rela

tive

8-O

H(d

)G e

xpre

ssio

n(fo

ld in

duct

ion

/WT)

C H S.C. P.C.

A8-OH(d)G,

GFAP, Hoechst

8-OH(d)G

8-OH(d)G, GFAP,

Hoechst

8-OH(d)G

8-OH(d)G, GFAP,

Hoechst

8-OH(d)G

8-OH(d)G, GFAP,

Hoechst

8-OH(d)G

DC

p < 0.001

p < 0.05p < 0.001

n.s.

Glul Glul x Alb-Cre+

Glul Glul x Alb-Cre+

Glul Glul x Alb-Cre+

Glul Glul x Alb-Cre+

GlulGlul x Alb-Cre+

Fig. 4. Liver-specific deletion of GS induces RNA oxidation in mouse brain.(A–D) Survey of 8-OH(d)G immunoreactivity in brain tissue from Glulfl/fl (Left)and Glulfl/fl × Alb-Cre+ (Right) mice is presented in the absence (Lower) orpresence (Upper) of costaining with GFAP (green) and Hoechst 34580 (blue).One representative of the cerebellum (A), hippocampus (B), somatosensorycortex (C), and piriform cortex (D) of n = 5 is demonstrated. (Scale bars: A,100 μm; B–D, 200 μm.) (E) Samples harvested from brain sections (from left toright: cerebellum, hippocampus, somatosensory cortex, and piriform cortex) ofGlulfl/fl and Glulfl/fl × Alb-Cre+ mice were tested for RNA oxidation expressionusing Northwestern blotting [n = 6 for Glulfl/fl mice, n = 9 (cerebellum, hip-pocampus, and piriform cortex) and n = 10 (somatosensory cortex) forGlulfl/fl ×Alb-Cre+ mice].

Qvartskhava et al. PNAS | April 28, 2015 | vol. 112 | no. 17 | 5523

PHYS

IOLO

GY

Dow

nloa

ded

by g

uest

on

July

15,

202

0

Glulfl/fl × Alb-Cre+ mice compared with WT mice was observed(Fig. 6A). Consistently, the overall distance traveled by Glulfl/fl ×Alb-Cre+ mice was increased compared with Glulfl/fl mice (Fig.6B). Furthermore, using the O-Maze test, Glulfl/fl × Alb-Cre+ micespent more time in the open arms compared with control ani-mals (Fig. 6C). In turn, Glulfl/fl × Alb-Cre+ mice spent less time inthe protected arms than Glulfl/fl mice (Fig. 6D). Taken together,these data indicate that chronic hyperammonemia results inbehavioral abnormalities. Furthermore, when we monitored theWT and Glulfl/fl × Alb-Cre+ mice over time, we observed slightlyreduced survival of GS-deficient mice compared with WT con-trols (Fig. S6), indicating that chronic ammonia exposure mayreduce the life span in vivo. Taken together, these data indicatethat deletion of GS in the liver results in hyperammonemia,phenotypic changes in brain tissue, and behavioral abnormalities.

DiscussionAccording to the present observations, lack of hepatic GS is fol-lowed by a marked increase of plasma ammonia levels. HepaticGS is selectively expressed in a small population of perivenoushepatocytes (4–6, 34, 35). This expression pattern depends onWnt-mediated activation of the transcriptional coactivatorβ-catenin (36, 37), which is also involved in hepatocyte pro-liferation (38–40). Consequently, liver-specific β-catenin–deficientmice exhibit defective GS activity in the liver (36). In our study,the severe hyperammonemia of the Glulfl/fl × Alb-Cre+ micehighlights the importance of this enzyme in a small population ofperivenous hepatocytes (so-called “perivenous scavenger cells”)as a high-affinity ammonia scavenger, which prevents ammoniaspillover from the periportal urea-synthesizing compartment intothe systemic circulation. The present observations show the in vivorelevance of a hypothesis postulated almost three decades ago (1,2, 24). As illustrated in Figs. 1C and 2C, the enzyme is specificallyexpressed in a small subpopulation of perivenous hepatocytes andis completely absent in the remainder of liver parenchyma. Thus,expression of the enzyme in this small population of hepatocytes,which amounts to only 6–7% of all liver parenchymal cells (5), ispivotal for the maintenance of nontoxic ammonia levels in theorganism. The hyperammonemia in hepatic failure is thus not

simply the result of impaired urea formation but may, in largepart, result from an impaired ability of the perivenous cells toincorporate ammonia into Gln. Apart from perivenous hepato-cytes, several other cell types, including skeletal muscle, can takeup ammonia, and thus contribute to ammonia detoxification (41).However, extrahepatic GS, which is intact in Glulfl/fl × Alb-Cre+

mice, is apparently not sufficient to prevent hyperammonemia inarterial blood and to compensate for the lack of GS in the liver.The safeguard of perivenous ammonia detoxification allows the

periportal hepatocytes to generate ammonia by Gln hydrolysis as asubstrate for periportal urea synthesis, which is a low-affinity sys-tem for ammonia detoxification (1). The ammonia generated byhepatic glutaminase, which is expressed in periportal hepatocytes,adds to the ammonia generated in the gastrointestinal tract. Suchhigh ammonia concentrations are apparently tolerated by hepa-tocytes but are toxic to the brain (42). Arterial ammonia readilycrosses the blood–brain barrier, thus entering cerebral tissue (43).In the brain, ammonia leads to Gln formation, astrocyte swelling,and generation of reactive oxygen and nitrogen species. Astrocyteswelling and oxidative/nitrosative stress affect signal transduction,neurotransmission, synaptic plasticity, and oscillatory networks inthe brain (42). The present observations revealed enhanced PTNand RNA oxidation in brain regions of Glulfl/fl × Alb-Cre+ mice,whose function is compromised in HE. Functions affected in HEinclude memory formation in the hippocampus (44); processing ofsensory stimuli, such as perception, in the somatosensory cortex(45); and motor functions in the cerebellum (46). Interestingly, inthe cerebellum of mice lacking hepatic GS, we identified excep-tionally high levels of oxidized RNA in Purkinje cells, which criti-cally control motor functions (30). In line with deranged cerebralmotor control, the present study identified increased locomotion inmice lacking hepatic GS with liver damage and/or cholestasis due tobile duct ligation (47, 48), portocaval shunting (49), or portal veinligation (14). This difference may suggest a role of impaired livermetabolism other than ammonia detoxification for decreased lo-comotion in these HE animal models. One may also speculate that

A B

C D

Fig. 6. Glulfl/fl × Alb-Cre+ mice show behavioral abnormalities comparedwith WT mice. (A and B) Glulfl/fl and Glulfl/fl × Alb-Cre+ mice were transferredinto a light barrier-equipped cage and allowed to adapt to the new envi-ronment for 24 h prior to starting the measurement. Total activity time (Left,n = 22 for Glulfl/fl mice and n = 23 for Glulfl/fl × Alb-Cre+ mice) and counts(Right, n = 23 for Glulfl/fl mice and n = 24 for Glulfl/fl × Alb-Cre+ mice) (A) anddistance traveled (n = 21 for Glulfl/fl mice and n = 22 for Glulfl/fl × Alb-Cre+

mice) (B) were determined after 24 h. Animals were monitored using theO-Maze test. Time (Left) and visits (Right) in the open arms (C) and time(Left) and visits (Right) in the closed arms (D) are presented (n = 10 for Glulfl/fl

mice and n = 11 for Glulfl/fl × Alb-Cre+ mice).

A

B

Expr

essi

on(fo

ld in

duct

ion/

Sdh

a)

Iba1 Cd140.001

0.004

0.016

0.064

0.256

1.024C

n.s.

n.s.

n.s.

Expr

essi

on(fo

ld in

duct

ion/

Sdh

a)

Il1β Il6

n.s.

n.s.

0.00001

0.00004

0.00016

0.00064

0.00256

0.01024

0.04096

Tnfα

GlulGlul x Alb-Cre+

GlulGlul x Alb-Cre+

Glul Glul x Alb-Cre+

Iba1

Fig. 5. Microglia activation marker and proinflammatory cytokine mRNAexpression in the cerebral cortex of WT and Glulfl/fl × Alb-Cre+ mice.(A) Detection of Iba1 in mouse cerebral cortex by wide-field fluorescencemicroscopy. One representative immunofluorescence analysis of three isshown. (Scale bars: 200 μm; Inset, 20 μm.) mRNA expression levels in mousecerebral cortex by RT-PCR of Iba1 and CD14 (B, n = 3) or IL-1β, I-L6, or TNF-α(C, n = 8 for Glulfl/fl mice and n = 6 of for Glulfl/fl × Alb-Cre+ mice) are shown.mRNA expression levels of Iba1, CD14, IL-1β, IL-6, or TNF-α were normalizedto succinate dehydrogenase complex subunit A (SDHA) mRNA levels.

5524 | www.pnas.org/cgi/doi/10.1073/pnas.1423968112 Qvartskhava et al.

Dow

nloa

ded

by g

uest

on

July

15,

202

0

chronic hyperammonemia from birth in Glulfl/fl × Alb-Cre+

mice may trigger so far unknown adaptations that are not presentin other experimental settings or in patients with liver failure.However, that chronic hyperammonemia in Glulfl/fl × Alb-Cre+

mice does not promote anxious behavior is consistent with findingsobtained from bile duct-ligated rats (47) or rats after portocavalanastomosis (50). Interestingly, oxidative stress markers were notelevated in the piriform cortex of liver-specific GS KO mice.Presently, the underlying reason, as well as the functional signifi-cance of this finding, remains enigmatic and needs to be exploredin further studies.Our findings also demonstrate that chronic hyperammonemia

in this mouse model is not sufficient to activate microglia in thecerebral cortex, which contrasts with findings on mouse brain afteracute liver failure due to liver ischemia or intoxication (51) orfindings on human brain tissue from cirrhotic patients with HE(32, 52). These discrepancies may be explained by species dif-ferences, allowance for compensatory mechanisms in chronichyperammonemia, and/or the complexity or severity of thepathophysiological setting in humans with liver cirrhosis. There-fore, other factors, such as systemic inflammation in acute liverfailure (51) and HE (53), might contribute to cerebral microgliaactivation in HE (32, 52). Indeed, a recent transcriptome analysison postmortem human brain samples indicated that systemic in-flammation may trigger altered cerebral expression of genes in-volved in inflammation in patients with liver cirrhosis and HE(52). Furthermore, our results demonstrate that cerebral synthesisof proinflammatory cytokines as observed in some animal modelsfor HE is not solely explained by hyperammonemia (16, 31), inline with previous data (13, 14). Consistent with the presentfindings, data derived from postmortem human brain samplesconsistently show no up-regulation of proinflammatory cytokinemRNA or protein in the brain tissue of patients with liver cirrhosisand HE (32, 52, 54).As shown in this paper, expression of functional GS in the

perivenous hepatic scavenger cell population is critical forammonia homeostasis in vivo. Liver cirrhosis in animal modelsand man and hepatocellular carcinoma are associated withstrongly reduced GS activity, followed by hyperammonemia (2,55–57). In cirrhosis with collaterals, down-regulation of GS inscavenger cells may be related to portocaval shunting (3, 58,59). Here, it should be noted that glutaminase activity in thecirrhotic liver increases four- to fivefold to compensate for amarked reduction of the capacity for urea synthesis, therebymaintaining a life-compatible urea cycle flux (55). Clinicalconditions such as sepsis (21) or portocaval anastomosis (58)are associated with reduced GS activity as a consequence of ascavenger cell defect, although flux through periportal ureasynthesis is not affected.The crucial role of hepatic GS for the maintenance of am-

monia homeostasis as shown in the present study is also sup-ported by a recent study using muscle-specific GS KO mice,where arterial ammonia levels were only marginally elevatedcompared with Glulfl/fl mice (60). Here, it was concluded thatskeletal muscle may only contribute about 3% of daily whole-bodyammonia detoxification. Moreover, our study shows no compen-satory up-regulation of GS in skeletal muscle in Glulfl/fl × Alb-Cre+

mice. Notably, the chronic hyperammonemia from birth in theabsence of liver damage in Glulfl/fl × Alb-Cre+ mice may re-semble hyperammonemia in children with inborn urea cycle disor-ders, who also exhibit cognitive and motor deficits (22). Thesedisorders emphasize the importance of the urea cycle for ammoniaclearance. Accordingly, both the high-affinity and low-affinity sys-tems represent two independent primary systems in the liver forammonia removal.In conclusion, the present study shows that lack of GS activity

in a small subpopulation of perivenous hepatocytes leads tohyperammonemia and cerebral ammonia intoxication.

Materials and MethodsAnimals. Gene-targeted mice lacking functional hepatic GS were generatedby inserting a loxP site, together with a flip-recombinase target flankedneomycin resistance cassette, downstream of the 3′ end of exon 3 and asingle loxP site between exon 1a and exon 1b (GenOway; Fig. 1A). Resultingmice were crossed with deleter mice expressing the flippase recombinase toremove the resistance cassette. Glulfl/fl mice were backcrossed to C57BL/6Jbackground more than 10 times. Next, mice were crossed to a mouseexpressing the Cre recombinase under the albumin promotor (25), resultingin liver-specific deletion of GS. Animals appeared viable and fertile, andshowed no signs of distress. Animals were kept under specific pathogen-freeconditions. Age- and sex-matched animals weighing over 20 g were used forexperiments. Most experiments were carried out with littermate controls ofmice aged between 8 and 12 wk. For determination of locomotor activity,animals between 3 and 75 wk of age were transferred into a light barrier-equipped cage monitored by a computer system (ActiMot/MoTil; TSE-Systems). Animals were allowed to adapt to the new environment for 24 h priorto starting the measurement for another 24 h. All animal experiments werereviewed and approved by the appropriate authorities, and were per-formed in accordance with the German animal protection law (Landesamtfür Natur, Umwelt und Verbraucherschutz Recklinghausen, RegierungspräsidiumTübingen, Tierveruchsanlage Düsseldorf). Genotyping and the O-Maze testof mice were performed as described in SI Materials and Methods.

Tissue Sampling. Detailed information on tissue sampling is provided in SIMaterials and Methods. The cerebral cortex was further dissected due tovisible morphological characteristics (61).

Video-Tracking. For data acquisition, animals were video-tracked by a 302050-SW-KIT-2-CAM camera (TSE-Systems) at a resolution of 0.62–0.72 pixel andanalyzed using the tracking software VideoMot2 (TSE-Systems).

Western Blot Analysis. Western blot analysis was carried out as described in SIMaterials and Methods and previously (11).

Northwestern Blot Analysis. Northwestern blot analysis was carried out asdescribed in SI Materials and Methods and previously (10).

Immunofluorescence Analysis. Immunofluorescence analysis was carried out asdescribed in SI Materials and Methods and previously (10).

GS Activity Assay. GS activity was measured according to the γ-glutamyltransferase reaction as described recently (21) and in SI Materials andMethods.

Real-Time PCR.Quantitative real-time PCR is described in detail in SI Materialsand Methods, including PCR primer sequences.

Serum Liver Enzyme Activity. Activity of Ala aminotransferase and Asp ami-notransferase was measured in serum using a serum multiple biochemicalanalyzer (Ektachem DTSCII; Johnson & Johnson).

Quantification of Urea Nitrogen in Serum and Urine. Urea nitrogen wasquantified using a Spotchem analyzer (Axonlab). Serum samples were pre-diluted 1:5 and urine samples were prediluted 1:50 in PBS (Panbiotech).

Blood Ammonia Determination. Blood ammonia was measured within 3 minafter blood sampling from the right heart ventricle using an Ammonia CheckerII (Daiichi Kagaku Co. Ltd., distributed by Nobis Labordiagnostica GmbH).

Statistical Analysis. Data are provided as means ± SEM, and n represents thenumber of animals tested. All data were tested for significance using apaired or unpaired Student’s t test or one-way ANOVA. Only results with P <0.05 were considered statistically significant.

ACKNOWLEDGMENTS. This study was supported by the Deutsche Forschungs-gemeinschaft through Collaborative Research Centers Grant Sonderfor-schungsbereich (SFB) 575 “Experimental Hepatology” (Düsseldorf) and GrantSFB 974 “Communication and Systems Relevance of Liver Injury and Re-generation” (Düsseldorf), the Alexander von Humboldt Foundation (GrantSKA2010), and the in vivo investigations in metabolic pathomechanismsand diseases in Düsseldorf Graduate School of the Heinrich Heine Univer-sity of Düsseldorf.

Qvartskhava et al. PNAS | April 28, 2015 | vol. 112 | no. 17 | 5525

PHYS

IOLO

GY

Dow

nloa

ded

by g

uest

on

July

15,

202

0

1. Häussinger D (1983) Hepatocyte heterogeneity in glutamine and ammonia metabo-lism and the role of an intercellular glutamine cycle during ureogenesis in perfusedrat liver. Eur J Biochem 133(2):269–275.

2. Häussinger D (1990) Nitrogen metabolism in liver: Structural and functional organi-zation and physiological relevance. Biochem J 267(2):281–290.

3. Häussinger D, Lamers WH, Moorman AF (1992) Hepatocyte heterogeneity in themetabolism of amino acids and ammonia. Enzyme 46(1-3):72–93.

4. Häussinger D, Schliess F (2007) Glutamine metabolism and signaling in the liver. FrontBiosci 12:371–391.

5. Gebhardt R, Mecke D (1983) Heterogeneous distribution of glutamine synthetaseamong rat liver parenchymal cells in situ and in primary culture. EMBO J 2(4):567–570.

6. Gaasbeek Janzen JW, et al. (1984) Immunohistochemical localization of carbamoyl-phosphate synthetase (ammonia) in adult rat liver; evidence for a heterogeneousdistribution. J Histochem Cytochem 32(6):557–564.

7. Häussinger D, Sies H (2013) Hepatic encephalopathy: Clinical aspects and pathoge-netic concept. Arch Biochem Biophys 536(2):97–100.

8. Norenberg MD, et al. (1991) Ammonia-induced astrocyte swelling in primary culture.Neurochem Res 16(7):833–836.

9. Häussinger D, Schliess F (2008) Pathogenetic mechanisms of hepatic encephalopathy.Gut 57(8):1156–1165.

10. Görg B, et al. (2008) Ammonia induces RNA oxidation in cultured astrocytes and brainin vivo. Hepatology 48(2):567–579.

11. Schliess F, et al. (2002) Ammonia induces MK-801-sensitive nitration and phosphor-ylation of protein tyrosine residues in rat astrocytes. FASEB J 16(7):739–741.

12. Roselló DM, et al. (2008) Oxidative stress and hippocampus in a low-grade hepatic en-cephalopathy model: Protective effects of curcumin. Hepatol Res 38(11):1148–1153.

13. Begega A, et al. (2010) Portal hypertension in 18-month-old rats: Memory deficits andbrain metabolic activity. Physiol Behav 100(2):135–142.

14. Brück J, et al. (2011) Locomotor impairment and cerebrocortical oxidative stress inportal vein ligated rats in vivo. J Hepatol 54(2):251–257.

15. Dhanda S, Kaur S, Sandhir R (2013) Preventive effect of N-acetyl-L-cysteine on oxi-dative stress and cognitive impairment in hepatic encephalopathy following bile ductligation. Free Radic Biol Med 56:204–215.

16. Rodrigo R, et al. (2010) Hyperammonemia induces neuroinflammation that contrib-utes to cognitive impairment in rats with hepatic encephalopathy. Gastroenterology139(2):675–684.

17. Görg B, et al. (2010) Oxidative stress markers in the brain of patients with cirrhosisand hepatic encephalopathy. Hepatology 52(1):256–265.

18. Rangroo Thrane V, et al. (2013) Ammonia triggers neuronal disinhibition and seizuresby impairing astrocyte potassium buffering. Nat Med 19(12):1643–1648.

19. Seifert G, Schilling K, Steinhäuser C (2006) Astrocyte dysfunction in neurologicaldisorders: A molecular perspective. Nat Rev Neurosci 7(3):194–206.

20. Häussinger D, Sies H, Gerok W (1985) Functional hepatocyte heterogeneity in am-monia metabolism. The intercellular glutamine cycle. J Hepatol 1(1):3–14.

21. Görg B, Wettstein M, Metzger S, Schliess F, Häussinger D (2005) Lipopolysaccharide-induced tyrosine nitration and inactivation of hepatic glutamine synthetase in therat. Hepatology 41(5):1065–1073.

22. Gropman AL, Batshaw ML (2004) Cognitive outcome in urea cycle disorders. MolGenet Metab 81(Suppl 1):S58–S62.

23. Häberle J, et al. (2005) Congenital glutamine deficiency with glutamine synthetasemutations. N Engl J Med 353(18):1926–1933.

24. Häussinger D, Gerok W (1984) Hepatocyte heterogeneity in ammonia metabolism:Impairment of glutamine synthesis in CCl4 induced liver cell necrosis with no effect onurea synthesis. Chem Biol Interact 48(2):191–194.

25. Postic C, et al. (1999) Dual roles for glucokinase in glucose homeostasis as determinedby liver and pancreatic beta cell-specific gene knock-outs using Cre recombinase.J Biol Chem 274(1):305–315.

26. Bode JG, Peters-Regehr T, Kubitz R, Häussinger D (2000) Expression of glutaminesynthetase in macrophages. J Histochem Cytochem 48(3):415–422.

27. Kuo FC, Hwu WL, Valle D, Darnell JE, Jr (1991) Colocalization in pericentral hepato-cytes in adult mice and similarity in developmental expression pattern of ornithineaminotransferase and glutamine synthetase mRNA. Proc Natl Acad Sci USA 88(21):9468–9472.

28. Weiner ID, Miller RT, Verlander JW (2003) Localization of the ammonium trans-porters, Rh B glycoprotein and Rh C glycoprotein, in the mouse liver. Gastroenter-ology 124(5):1432–1440.

29. Görg B, et al. (2007) Reversible inhibition of mammalian glutamine synthetase bytyrosine nitration. FEBS Lett 581(1):84–90.

30. Barski J, et al. (2003) Calbindin in cerebellar Purkinje cells is a critical determinant ofthe precision of motor coordination. J Neurosci 23(8):3469–3477.

31. Jiang W, Desjardins P, Butterworth RF (2009) Direct evidence for central proin-flammatory mechanisms in rats with experimental acute liver failure: Protective effectof hypothermia. J Cereb Blood Flow Metab 29(5):944–952.

32. Zemtsova I, et al. (2011) Microglia activation in hepatic encephalopathy in rats and

humans. Hepatology 54(1):204–215.33. Glass CK, Saijo K, Winner B, Marchetto MC, Gage FH (2010) Mechanisms underlying

inflammation in neurodegeneration. Cell 140(6):918–934.34. Burke ZD, et al. (2009) Liver zonation occurs through a beta-catenin-dependent,

c-Myc-independent mechanism. Gastroenterology 136(7):2316–2324. e1–e3.35. Clinkenbeard EL, Butler JE, Spear BT (2012) Pericentral activity of alpha-fetoprotein

enhancer 3 and glutamine synthetase upstream enhancer in the adult liver are reg-

ulated by β-catenin in mice. Hepatology 56(5):1892–1901.36. Sekine S, Lan BY-A, Bedolli M, Feng S, Hebrok M (2006) Liver-specific loss of beta-

catenin blocks glutamine synthesis pathway activity and cytochrome p450 expression

in mice. Hepatology 43(4):817–825.37. Yang J, et al. (2014) β-catenin signaling in murine liver zonation and regeneration: A

Wnt-Wnt situation! Hepatology 60(3):964–976.38. Sekine S, Gutiérrez PJA, Lan BY-A, Feng S, Hebrok M (2007) Liver-specific loss of beta-

catenin results in delayed hepatocyte proliferation after partial hepatectomy. Hep-

atology 45(2):361–368.39. Tan X, Behari J, Cieply B, Michalopoulos GK, Monga SPS (2006) Conditional deletion

of beta-catenin reveals its role in liver growth and regeneration. Gastroenterology

131(5):1561–1572.40. Nejak-Bowen KN, et al. (2010) Accelerated liver regeneration and hepatocarcino-

genesis in mice overexpressing serine-45 mutant beta-catenin. Hepatology 51(5):

1603–1613.41. Dam G, Ott P, Aagaard NK, Vilstrup H (2013) Branched-chain amino acids and muscle

ammonia detoxification in cirrhosis. Metab Brain Dis 28(2):217–220.42. Butz M, May ES, Häussinger D, Schnitzler A (2013) The slowed brain: Cortical oscil-

latory activity in hepatic encephalopathy. Arch Biochem Biophys 536(2):197–203.43. Sørensen M (2013) Update on cerebral uptake of blood ammonia. Metab Brain Dis

28(2):155–159.44. Weissenborn K, Heidenreich S, Giewekemeyer K, Rückert N, Hecker H (2003) Memory

function in early hepatic encephalopathy. J Hepatol 39(3):320–325.45. Brenner M, et al. (2015) Patients with manifest hepatic encephalopathy can reveal

impaired thermal perception. Acta Neurol Scand, 10.1111/ane.12376.46. Butz M, et al. (2010) Motor impairment in liver cirrhosis without and with minimal

hepatic encephalopathy. Acta Neurol Scand 122(1):27–35.47. Leke R, et al. (2012) Impairment of the organization of locomotor and exploratory

behaviors in bile duct-ligated rats. PLoS ONE 7(5):e36322.48. Magen I, et al. (2009) Cannabidiol ameliorates cognitive and motor impairments in

mice with bile duct ligation. J Hepatol 51(3):528–534.49. Cauli O, Llansola M, Erceg S, Felipo V (2006) Hypolocomotion in rats with chronic liver

failure is due to increased glutamate and activation of metabotropic glutamate re-

ceptors in substantia nigra. J Hepatol 45(5):654–661.50. Sergeeva OA, et al. (2005) Deficits in cortico-striatal synaptic plasticity and behavioral

habituation in rats with portacaval anastomosis. Neuroscience 134(4):1091–1098.51. Butterworth RF (2011) Neuroinflammation in acute liver failure: Mechanisms and

novel therapeutic targets. Neurochem Int 59(6):830–836.52. Görg B, Bidmon H-J, Häussinger D (2013) Gene expression profiling in the cerebral

cortex of patients with cirrhosis with and without hepatic encephalopathy. Hep-

atology 57(6):2436–2447.53. Coltart I, Tranah TH, Shawcross DL (2013) Inflammation and hepatic encephalopathy.

Arch Biochem Biophys 536(2):189–196.54. Dennis CV, et al. (2014) Microglial proliferation in the brain of chronic alcoholics with

hepatic encephalopathy. Metab Brain Dis 29(4):1027–1039.55. Kaiser S, Gerok W, Häussinger D (1988) Ammonia and glutamine metabolism in hu-

man liver slices: New aspects on the pathogenesis of hyperammonaemia in chronic

liver disease. Eur J Clin Invest 18(5):535–542.56. Matsuno T, Goto I (1992) Glutaminase and glutamine synthetase activities in human

cirrhotic liver and hepatocellular carcinoma. Cancer Res 52(5):1192–1194.57. Gebhardt R, Reichen J (1994) Changes in distribution and activity of glutamine syn-

thetase in carbon tetrachloride-induced cirrhosis in the rat: Potential role in hyper-

ammonemia. Hepatology 20(3):684–691.58. Kennan AL (1964) Glutamine synthetase activity in rat liver after portacaval shunt.

Endocrinology 74:805–806.59. Girard G, Butterworth RF (1992) Effect of portacaval anastomosis on glutamine syn-

thetase activities in liver, brain, and skeletal muscle. Dig Dis Sci 37(7):1121–1126.60. He Y, et al. (2010) Glutamine synthetase in muscle is required for glutamine pro-

duction during fasting and extrahepatic ammonia detoxification. J Biol Chem 285(13):

9516–9524.61. Hof PR, Young WG, Bloom FL, Belichenko PV, Celio M (2000) Comparative Cytoarchi-

tectonic Atlas of the C57BL/6 and 129/Sv Mouse Brains (Elsevier, Amsterdam).

5526 | www.pnas.org/cgi/doi/10.1073/pnas.1423968112 Qvartskhava et al.

Dow

nloa

ded

by g

uest

on

July

15,

202

0